Angle Estimation Methods Of Array Signal Based On Sparse Signal Reconstruction And Distributed Sources

Angle estimation technique in Array signal processing has great prospects for application.It has been an important technique in many applications,i.e.,Space-Division Multiple Access in smart antennas,electromagnetic countermeasure,acoustic location,etc.Experts and scholars from different countries have proposed a lot of methods of good performance over the past few decades.With the further study and wider application,however,people put forward higher requirements to the adaptability of the angle estimation methods to the complex signal environment.Practical applications require the adaptability of low signal-noise-ratio and few snapshots,the robustness to the model mismatch,the ability of angle parameter estimation for multiple distributed sources,and even the applicability to direction finding with nonuniform linear array.With the practical requirement on the low complexity of the algorithms taken into account,our research focuses on satisfying these requirements and studies on joint angle estimation for bistatic Multiple-Input Multiple-Output(MIMO)radar,direction of arrival(DOA)estimation under nonuniform noise and nominal DOA estimation for distributed sources with its application on special array,respectively.Besides,we propose a few novel methods which play an active role in promoting the engineering application of angle estimation technique.The main researches and contributions are as follows.1.Conventional bistatic MIMO radar has no transmit coherent processing gain,so the Signal-to-Noise-Ratio gain is low.For solving this problem,we propose multi-overlapped-transmit-subarray configured bistatic MIMO radar system.The transmit subarray which works on the phased array mode makes the proposed bistatic MIMO radar enjoy both of the waveform diversity gain and transmit coherent processing gain so that it could better accommodate the situations of low signal-noise-ratio and few snapshots.We propose an unitary Estimation Signal Parameters via Rotational Invariance Techniques(ESPRIT)method for joint DOD and DOA estimation and a least-square based auto-pairing method for angle estimates.The unitary ESPRIT method not only doubles the effective measurement data,but also greatly improves computation efficiency by transforming the complex-valued computation into the real-valued computation.As a result,the proposed scheme enables the bistatic MIMO radar to efficiently obtain the joint angle estimation under low signal-noise-ratio and few snapshots,which is significant in distant target detection,disguised emitter signal interception and real-time parameter estimation for the signal which rapidly changes with time.2.In practice,both of the nonuniformity of the sensor noise power and the quantization error of over-complete discretized angle set can cause array measurement model error,which leads to the deteriorated performance on DOA estimation.To overcome this problem,we propose a sparsity-inducing method of DOA estimation based on partial virtual array output.By exploiting the joint sparsity of signal power and off-grid difference,a sparse signal reconstruction problem with off-grid parameter is established on basis of the signal model of the partial virtual array output under nonuniform noise.We propose an off-grid variational sparse Bayesian learning based off-grid DOA estimation method,by employing the over-complete dictionary constructed according to the manifold matrix including virtual array steering vector with off-grid parameter.The proposed method models the estimation error of the weighted partial covariance vector as white Gaussian noise vector so that nonuniform noise variance estimation can be avoided and the effect of the nonuniform noise on the algorithm can be alleviated.A three-level Bayesian model is established with an almost Jeffrey prior at the second level of hierarchy,which strongly induces the sparsity of signal power distribution and improve the precision of the estimation.By adaptively updating the off-grid parameter and adjusting the over-complete dictionary,the proposed method could reduce the off-grid error caused by the quantization error of the over-complete discretized angle set,which leads to the improved precision of DOA estimation.Experimental results show that the proposed DOA estimation method is robust to the nonuniform noise and the computation grid interval,and it is able to accommodate the condition that there exists both of the model error mentioned above and the factors under complex signal environment including low signal-noise-ratio,few snapshots,unknown source number and multiple correlated signals.This finally improves the adaptability of the proposed DOA estimation method to the complex environment.3.In practice,some factors,such as,multipath propagation,local scattering,etc.can cause angle spread of the sources,which deteriorates the performance of point source model based DOA estimation method.For solving this problem,we propose a nominal DOA estimation method of multiple distributed sources using a nested array.A multiplicative noise model is established for multiple distributed sources based on nested array.According to the difference coarray of the nested array,we establish the matching criterion on basis of the redundancy averaged covariance terms with coherence loss function and propose two nominal DOA estimation methods for multiple distributed sources.According to the annihilating property of the redundancy averaged covariance terms,an annihilation filter based estimation method is proposed.Furthermore,an annihilation filter based estimation method with structured low rank approximation is proposed,in order to enhance the robustness to the low signal-noise-ratio and few snapshots.By exploiting the property that the difference coarray of the two-level nested array is similar to a uniform linear array with a larger aperture and more sensors,the proposed methods can be applied to the underdetermined DOA estimation where the number of distributed sources is larger than that of the physical sensors.It is worth noting that the proposed methods can obtain angle estimates without any spectrum search,which reduces the computation burden,and besides,it does not require the prior knowledge of the angle spread parameters.